CN116504330A - Pollutant concentration inversion method and device, electronic equipment and readable storage medium - Google Patents
Pollutant concentration inversion method and device, electronic equipment and readable storage medium Download PDFInfo
- Publication number
- CN116504330A CN116504330A CN202310770616.6A CN202310770616A CN116504330A CN 116504330 A CN116504330 A CN 116504330A CN 202310770616 A CN202310770616 A CN 202310770616A CN 116504330 A CN116504330 A CN 116504330A
- Authority
- CN
- China
- Prior art keywords
- data
- concentration inversion
- concentration
- inversion
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 239000003344 environmental pollutant Substances 0.000 title claims abstract description 82
- 231100000719 pollutant Toxicity 0.000 title claims abstract description 82
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000012549 training Methods 0.000 claims abstract description 24
- 239000011159 matrix material Substances 0.000 claims description 69
- 230000005855 radiation Effects 0.000 claims description 29
- 239000000356 contaminant Substances 0.000 claims description 27
- 238000012417 linear regression Methods 0.000 claims description 17
- 238000012545 processing Methods 0.000 claims description 17
- 230000009467 reduction Effects 0.000 claims description 17
- 238000013528 artificial neural network Methods 0.000 claims description 10
- 230000004927 fusion Effects 0.000 claims description 8
- 238000012706 support-vector machine Methods 0.000 claims description 7
- CBENFWSGALASAD-UHFFFAOYSA-N Ozone Chemical compound [O-][O+]=O CBENFWSGALASAD-UHFFFAOYSA-N 0.000 description 46
- 238000012544 monitoring process Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 8
- 230000008901 benefit Effects 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 6
- 239000007789 gas Substances 0.000 description 6
- 238000004422 calculation algorithm Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 238000010521 absorption reaction Methods 0.000 description 4
- 230000004913 activation Effects 0.000 description 4
- 230000008859 change Effects 0.000 description 4
- 238000004891 communication Methods 0.000 description 4
- 238000013136 deep learning model Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000004590 computer program Methods 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- 238000001658 differential optical absorption spectrophotometry Methods 0.000 description 3
- 230000003595 spectral effect Effects 0.000 description 3
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000007637 random forest analysis Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000012800 visualization Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 238000012952 Resampling Methods 0.000 description 1
- 238000004847 absorption spectroscopy Methods 0.000 description 1
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 1
- 230000008033 biological extinction Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000000802 evaporation-induced self-assembly Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 229910052760 oxygen Inorganic materials 0.000 description 1
- 239000001301 oxygen Substances 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 239000005436 troposphere Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/30—Prediction of properties of chemical compounds, compositions or mixtures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Tourism & Hospitality (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Crystallography & Structural Chemistry (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Chemical & Material Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Mathematical Physics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Image Processing (AREA)
- Image Generation (AREA)
- Nuclear Medicine (AREA)
- Semiconductor Memories (AREA)
Abstract
The invention provides a pollutant concentration inversion method, a pollutant concentration inversion device, electronic equipment and a readable storage medium, which comprise the following steps: acquiring a first data set corresponding to a region to be researched; according to the land coverage type in the first time-space data, determining a target concentration inversion model corresponding to each pixel point in each region to be researched from a concentration inversion model cluster obtained through pre-training; the concentration inversion model cluster comprises concentration inversion models corresponding to various land coverage types; and carrying out pollutant concentration inversion on the region to be researched based on the first data set through each target concentration inversion model to obtain a target concentration inversion result corresponding to the region to be researched. The invention can remarkably reduce the complexity of inversion of the pollutant concentration, can effectively improve the inversion efficiency of the pollutant concentration, and can remarkably improve the inversion precision of the pollutant concentration.
Description
Technical Field
The invention relates to the technical field of pollutant inversion, in particular to a pollutant concentration inversion method, a pollutant concentration inversion device, electronic equipment and a readable storage medium.
Background
Ozone is one of the atmospheric trace gases which has important influence on ecology, climate and environment, and the ozone in the atmosphere has active chemical reaction characteristics and strong radiation characteristics, and can directly influence global climate change and human living environment. The existing hyperspectral load ozone concentration inversion algorithm is mostly based on foreign satellite data, such as TOMS (Total Ozone Mapping Spectrometer ), DOAS (Differential Optical Absorption Spectroscopy, differential absorption spectroscopy technology) and the like, and has the problems of complex calculation process, low calculation efficiency when calculating a large amount of data, and the like.
Disclosure of Invention
In view of the above, the present invention aims to provide a method, an apparatus, an electronic device, and a readable storage medium for inverting a contaminant concentration, which can significantly reduce the complexity of the inversion of the contaminant concentration, effectively improve the inversion efficiency of the contaminant concentration, and significantly improve the inversion accuracy of the contaminant concentration.
In a first aspect, an embodiment of the present invention provides a method for inverting a concentration of a contaminant, including:
acquiring a first data set corresponding to a region to be researched; the first data set comprises first remote sensing image data, first ground observation data and first time-space data;
According to the land coverage type in the first time-space data, determining a target concentration inversion model corresponding to each pixel point in each region to be researched from a concentration inversion model cluster obtained through pre-training; the concentration inversion model cluster comprises concentration inversion models corresponding to various land coverage types;
and carrying out pollutant concentration inversion on the region to be researched based on the first data set through each target concentration inversion model to obtain a target concentration inversion result corresponding to the region to be researched.
In one embodiment, the target concentration inversion model includes a first stage combined sub-model and a second stage combined sub-model;
carrying out pollutant concentration inversion on the region to be researched based on the first data set through each target concentration inversion model to obtain a target concentration inversion result corresponding to the region to be researched, wherein the method comprises the following steps:
determining at least one initial concentration inversion result corresponding to the pixel point based on the first remote sensing image data, the first ground observation data and the first time-space data through the first stage combination sub-model; the first-stage combined sub-model comprises at least one inversion unit, wherein each inversion unit is used for determining an initial concentration inversion result corresponding to the pixel point;
Determining a target concentration inversion result corresponding to the pixel point based on the first remote sensing image data, the first ground observation data, the first time-space data and each initial concentration inversion result through the second-stage combined sub-model;
and obtaining the target concentration inversion result corresponding to the region to be researched based on the target concentration inversion result corresponding to each pixel point.
In one embodiment, the second stage combining sub-model includes a linear regression network, a feed forward neural network, and a fusion network, and the first time space data further includes ecology geographic type data and landform type data;
determining, by the second stage combined sub-model, a target concentration inversion result corresponding to the pixel point based on the first remote sensing image data, the first ground observation data, the first time-space data, and each of the initial concentration inversion results, including:
performing single-heat encoding processing on the ecological geography type data and the landform type data through the linear regression network to obtain a sparse matrix, and determining a first concentration inversion result corresponding to the pixel point according to the first remote sensing image data, the first ground observation data, the sparse matrix and each initial concentration inversion result;
Performing dimension reduction processing on the ecological geography type data and the landform type data through the feedforward neural network to obtain a dense matrix, and determining a second concentration inversion result corresponding to the pixel point according to the first remote sensing image data, the first ground observation data, the dense matrix and each initial concentration inversion result;
and weighting the first concentration inversion result and the second concentration inversion result through the fusion network to obtain a target concentration inversion result corresponding to the pixel point.
In one embodiment, the inversion unit is one or more of an XGBoost unit, a ridge regression unit, and a support vector machine unit.
In one embodiment, the training step of the concentration inversion model cluster includes:
acquiring a second data set; the second data set comprises second remote sensing image data, second ground observation data, second space-time data and ground observation station pollutant data;
dividing the second data set into a plurality of sub-data sets based on a land cover type in the second spatio-temporal data;
for each of the sub-data sets, determining a predicted concentration inversion result based on the second remote sensing image data, the second ground observation data, and the second spatiotemporal data in the sub-data set by an initial concentration inversion model;
Training the initial concentration inversion model based on the pollutant data of the ground observation station in the sub-data set and the predicted concentration inversion result to obtain a target concentration inversion model corresponding to the sub-data set;
and combining the target concentration inversion models corresponding to each sub-data set into a concentration inversion model cluster.
In one embodiment, the performing the inversion of the pollutant concentration on the region to be studied based on the first data set to obtain a target concentration inversion result corresponding to the region to be studied, further includes:
performing radiation calibration on the first remote sensing image data to obtain a calibrated radiation brightness matrix;
zero-equalizing the scaled radiation brightness matrix to obtain an equalized radiation brightness matrix;
determining a covariance matrix based on the scaled radiation brightness matrix and the averaged radiation brightness matrix, and performing data dimension reduction processing based on the eigenvalues and eigenvectors of the covariance matrix to obtain dimension reduced data;
and carrying out pollutant concentration inversion on the region to be researched based on the dimension-reduced data, the first ground observation data and the first time space data through each target concentration inversion model to obtain a target concentration inversion result corresponding to the region to be researched.
In one embodiment, after inverting the contaminant concentration of the region to be studied based on the first data set, to obtain a target concentration inversion result corresponding to the region to be studied, the method further includes:
superposing target grid data corresponding to the region to be researched and the target concentration inversion result to determine a pollutant concentration value corresponding to each grid or determine a pollutant concentration average value corresponding to each grid;
and sending the pollutant concentration value or the pollutant concentration mean value to a designated associated terminal so as to visually display the pollutant concentration value or the pollutant concentration mean value through a graphic user interface of the designated associated terminal.
In a second aspect, an embodiment of the present invention further provides a contaminant concentration inversion apparatus, including:
the data acquisition module is used for acquiring a first data set corresponding to the region to be researched; the first data set comprises first remote sensing image data, first ground observation data and first time-space data;
the model determining module is used for determining a target concentration inversion model corresponding to each pixel point in each region to be researched from concentration inversion model clusters obtained through pre-training according to the land coverage type in the first time-space data; the concentration inversion model cluster comprises concentration inversion models corresponding to various land coverage types;
And the concentration inversion module is used for inverting the concentration of the pollutant in the region to be researched based on the first data set through each target concentration inversion model to obtain a target concentration inversion result corresponding to the region to be researched.
In a third aspect, an embodiment of the present invention further provides an electronic device comprising a processor and a memory storing computer-executable instructions executable by the processor to implement the method of any one of the first aspects.
In a fourth aspect, embodiments of the present invention also provide a computer-readable storage medium storing computer-executable instructions which, when invoked and executed by a processor, cause the processor to implement the method of any one of the first aspects.
According to the pollutant concentration inversion method, the device, the electronic equipment and the readable storage medium, a first data set corresponding to an area to be researched is firstly obtained, the first data set comprises first remote sensing image data, first ground observation data and first time space data, then according to the land coverage type in the first time space data, a target concentration inversion model (comprising concentration inversion models corresponding to various land coverage types) corresponding to each pixel point in each area to be researched is determined from concentration inversion model clusters obtained through training in advance, finally pollutant concentration inversion can be carried out on the area to be researched based on the first data set through each target concentration inversion model, and a target concentration inversion result corresponding to the area to be researched is obtained. According to the method, aiming at the first remote sensing image data, the first ground observation data and the first time space data are combined, meanwhile, the pollutant concentration in the area to be researched is inverted by utilizing the corresponding target concentration inversion model in the concentration inversion model cluster based on the land coverage types in different first time space data, and inversion errors caused by the assumption of the ideal underlying surface by the traditional model are overcome, so that the complexity of inversion of the pollutant concentration can be remarkably reduced, the inversion efficiency of the pollutant concentration can be effectively improved, and the inversion precision of the pollutant concentration can be remarkably improved.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
In order to make the above objects, features and advantages of the present invention more comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the description of the embodiments or the prior art will be briefly described, and it is obvious that the drawings in the description below are some embodiments of the present invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of a method for inverting the concentration of a contaminant according to an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of a target inversion model according to an embodiment of the present invention;
FIG. 3 is a schematic view of an observation coverage area according to an embodiment of the present invention;
FIG. 4 is a schematic view of an ozone observation effect according to an embodiment of the present invention;
FIG. 5 is a schematic structural diagram of a pollutant concentration inversion apparatus according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described in conjunction with the embodiments, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Ozone is one of the atmospheric trace gases which has important influence on ecology, climate and environment, and the ozone in the atmosphere has active chemical reaction characteristics and strong radiation characteristics, and can directly influence global climate change and human living environment. In 1985, FARMAN et al first found a Antarctic ozone hole, and then found many miniature ozone low-value regions in Qinghai-Tibet plateau in China and other plateau regions of the world. Research on the cause of ozone holes and the mechanism of ozone change requires ozone monitoring data with high space-time resolution. Compared with a ground-based ozone observation instrument, the satellite ozone detector has obvious advantages in space-time aspect, and satellite ozone detection becomes an important means for monitoring global ozone changes.
A series of hyperspectral ultraviolet ozone detectors such as an oxygen monitoring instrument (GOME), an atmospheric scanning imaging absorption Spectrometer (SCIAMACHY), an Ozone Monitoring Instrument (OMI), a global ozone monitoring instrument of generation 2 (GNOME-2), an ozone imaging profile instrument (OMPS), a troposphere monitoring instrument and the like are used for global ozone change satellite detection.
2021, 9, 7, 11 and 01 minutes, five times higher 02 satellites were successfully launched in the Taiyuan satellite launch center. The high-resolution No. 02 star is also called hyperspectral observation satellite, is a service star in the project of long-term development in civil space infrastructure (2015-2025), provides domestic hyperspectral data for environmental protection main body services such as monitoring of atmospheric environment, water environment and ecological environment, and is used for industrial applications such as homeland resources, disaster prevention and reduction, agriculture, forestry and weather. The trace gas differential absorption spectrometer (EMI) becomes the only available hyperspectral pollution gas monitoring load after being in orbit, and the EMI has an ultraviolet hyperspectral detection means of 0.5nm, so that the monitoring of the global pollution gas can be covered on a single day.
The existing hyperspectral load ozone concentration inversion algorithm is mostly based on foreign satellite data, such as TOMS, DOAS and the like, and has the problems of complex calculation process, low calculation efficiency of a large amount of data and the like.
Based on the method, the device, the electronic equipment and the readable storage medium for inverting the pollutant concentration are provided, so that the complexity of inverting the pollutant concentration can be remarkably reduced, the inversion efficiency of the pollutant concentration can be effectively improved, and the inversion precision of the pollutant concentration can be remarkably improved.
For the sake of understanding the present embodiment, first, a detailed description will be given of a method for inverting the concentration of a contaminant disclosed in the present embodiment, where the contaminant may include O3, PM2.5, PM10, SO2, CO, NO2, etc., and preferably, the embodiment of the present invention may be used for inverting the concentration of ozone, referring to a schematic flow chart of a method for inverting the concentration of a contaminant shown in fig. 1, and the method mainly includes the following steps S102 to S106:
step S102, a first data set corresponding to a region to be researched is obtained; the first data set comprises first remote sensing image data, first ground observation data and first time space data, and the first data set is a data set used in a model application stage.
In one example, the first remote sensing image data selects GF5B/EMI (UV 2 wave band) data as original observation data, the spectral wavelength range of the UV2 wave band data is 302-403 nm, and the wavelength range data of 350-403 nm is selected for algorithm inversion because the spectral wavelength range is smaller than 350nm and is mainly O3 absorption band; the data involved in the inversion also includes relative azimuth (solar azimuth minus absolute value of satellite azimuth), solar zenith angle, satellite zenith angle and longitude and latitude data. In one example, the first ground observation data includes pressure, humidity, temperature, wind speed, boundary layer height. In one example, the first time-space data may include an ecology geographic type, a geomorphic type, and may further include land cover data.
Step S104, determining a target concentration inversion model corresponding to each pixel point in each region to be researched from concentration inversion model clusters obtained through pre-training according to the land coverage type in the first time-space data. The concentration inversion model cluster comprises concentration inversion models corresponding to various land coverage types. Optionally, land cover types are categorized by water, building, desert, forest land, grassland, farmland, and others.
In one embodiment, a land coverage type of each pixel point may be determined according to land coverage data in the first time-space data, and then a target concentration inversion model corresponding to the pixel point may be selected from a concentration inversion model cluster according to the land coverage type.
And S106, carrying out pollutant concentration inversion on the region to be researched based on the first data set through each target concentration inversion model to obtain a target concentration inversion result corresponding to the region to be researched. The target concentration inversion result may be a contaminant concentration value. In one embodiment, for each pixel point, the inversion result of the initial concentration of the pixel point may be inverted based on the first remote sensing image data, the first ground observation data and the first time space data corresponding to the pixel point according to the inversion model of the target concentration corresponding to the pixel point, and then the inversion result of the target concentration of the pixel point may be inverted based on the first remote sensing image data, the first ground observation data, the first time space data and the inversion result of the initial concentration corresponding to the pixel point. After the inversion result of the target concentration of each pixel point is determined, the inversion result of the target concentration corresponding to the region to be researched can be obtained.
According to the pollutant concentration inversion method provided by the embodiment of the invention, aiming at the first remote sensing image data, the first ground observation data and the first time space data are combined, meanwhile, the pollutant concentration in the area to be researched is inverted by utilizing the corresponding target concentration inversion model in the concentration inversion model cluster based on the land coverage types in different first time space data, and the inversion error caused by the assumption of the ideal underlying surface by the traditional model is overcome, so that the complexity of the pollutant concentration inversion can be remarkably reduced, the inversion efficiency of the pollutant concentration can be effectively improved, and the inversion precision of the pollutant concentration can be remarkably improved.
For ease of understanding, embodiments of the present invention provide a specific implementation of a contaminant concentration inversion method.
After the first data set is acquired, performing data dimension reduction processing on the first remote sensing image data in the first data set so as to facilitate the follow-up passing through each target concentration inversion model, and performing pollutant concentration inversion on the region to be researched based on the dimension reduced data, the first ground observation data and the first time space data to obtain a target concentration inversion result corresponding to the region to be researched. The embodiment of the invention provides a specific implementation mode of data dimension reduction processing, which is referred to as the following steps a to c:
And a step a, performing radiation calibration on the first remote sensing image data to obtain a calibrated radiation brightness matrix. In one embodiment, the selected first remote sensing image data (350-403 nm) is taken as a pixel brightness value, the first remote sensing image data is subjected to radiation calibration, and a calibration formula is used as follows:
；
wherein DN is pixel brightness value of the first remote sensing image data, radCaliCoeff is radiation calibration coefficient, and Radiance is radiation brightness after calibration.
The calibrated radiance data still has 572 spectrum channels, the data dimension is higher, the calculated amount is overlarge when the data is substituted into a deep learning model, and the data redundancy between adjacent wave bands is higher. The embodiment of the invention takes 10 GF5B/EMI data as an example, and 10 sample data are assumed to be counted in totalData points, each dimension represents the number of samples, latitude, longitude, and selected spectral channel. Because of the main spectrumThe trace is subjected to dimension reduction processing, so the data quantity is expressed asData points.
The data is organized into n rows and m columns of matrix X, i.e., a scaled radiance matrix, where n represents the data dimension, i.e., 572, and m represents the number of samples, i.e., 1176470.
。
And b, performing zero-averaging treatment on the calibrated radiation brightness matrix to obtain an average radiation brightness matrix. In one embodiment, each row of the scaled radiance matrix X is zero-averaged, i.e., the average of that row is subtracted, to obtain an averaged radiance matrix ：
。
And c, determining a covariance matrix based on the calibrated radiation brightness matrix and the averaged radiation brightness matrix, and performing data dimension reduction processing based on the eigenvalues and eigenvectors of the covariance matrix to obtain dimension reduced data. In one embodiment, see steps c1 to c4 below:
in step C1, the covariance matrix C may be calculated according to the following formula:
。
c is a symmetric matrix, the diagonal of which corresponds to the variance of each variable, and the j-th row, j, and j-th row, i, have the same column elements, representing the covariance of both i and j variables.
And C2, determining the eigenvalue and the corresponding eigenvector of the covariance matrix C. Covariance matrix C is a symmetric matrix, and real symmetric matrices in linear algebra have a series of very good properties: (1) The eigenvectors corresponding to different eigenvalues of the real symmetric matrix are necessarily orthogonal; (2) Assuming that the eigenvector λ is r in number, there are necessarily r linearly independent eigenvectors corresponding to λ, so these r eigenvectors can be orthogonalized in units.
From the above two, an n-row n-column real symmetric matrix must find n unit orthogonal eigenvectors, and let these n eigenvectors be e1, e2, & gt, en, form a matrix according to the columns: e= (E1, E2,) en. Then the covariance matrix C is concluded as follows:
；
Where Λ is a diagonal matrix, and its diagonal elements are eigenvalues (possibly duplicates) corresponding to each eigenvector.
In step C3, ET is denoted as a matrix P, i.e., p=et, where P is the matrix listed in rows after the eigenvectors of the covariance matrix are unitized, where each row is one eigenvector of C. The eigenvectors are arranged into a matrix according to the corresponding eigenvalue from top to bottom and the first k rows are taken to form a matrix Pk.
Step c4, y= PkX is the data after the dimension reduction to k dimension (i.e., the data after the dimension reduction). For example, in the embodiment of the present invention, k is set to 3, so the amount of data after dimension reduction becomes。
On the basis of the foregoing embodiment, the embodiment of the present invention provides a target concentration inversion model, where the target concentration inversion model provided in the embodiment of the present invention adopts a combined deep learning model, performs model training on the data after dimension reduction, and the model is divided into two major parts: (1) Performing initial estimation (i.e., initial concentration inversion result) on the pollutants by a first-stage combined sub-model, wherein the partial model consists of three deep learning models of XGBoost, ridge regression and a support vector machine based on Random Forest (RF); (2) Further inversion is performed on the initial estimate by combining sub-models in a second stage to obtain a more accurate inversion result, wherein the partial models jointly comprise a linear regression model and a feed-forward neural network model (Feedforward Neural Network, FNN).
Specifically, referring to a schematic structure diagram of a target inversion model shown in fig. 2, fig. 2 illustrates that the target concentration inversion model includes a first-stage combined sub-model and a second-stage combined sub-model, where the first-stage combined sub-model includes at least one inversion unit, each inversion unit is used for determining an initial concentration inversion result corresponding to a pixel point, and the inversion unit is one or more of an XGBoost unit, a ridge regression unit and a support vector machine unit.
With continued reference to fig. 2, fig. 2 illustrates that the second stage combining sub-model includes a linear regression network, a feedforward neural network, and a fusion network, the linear regression network includes a single thermal encoding layer and a linear regression layer, the linear regression network includes an embedding layer and a plurality of hidden layers, each including a fully connected layer (FC), an exponential linear unit activation function (eLU), a Batch Normalization (BN), and a dropout layer.
On the basis, the embodiment of the invention provides an implementation mode for carrying out pollutant concentration inversion on a region to be researched based on a first data set through each target concentration inversion model to obtain a target concentration inversion result corresponding to the region to be researched, which is described in the following steps 1 to 3:
And step 1, determining at least one initial concentration inversion result corresponding to the pixel point based on the first remote sensing image data, the first ground observation data and the first time space data through the first stage combined sub-model. In one embodiment, the first stage combined sub-model includes three units, such as a ridge regression unit, a support vector machine unit, and an XGBoost unit, where training and inversion are performed on the three units, respectively, and the initial concentration inversion result output by the three units is used as input data of the second stage combined sub-model.
For example, assuming that the contaminant is ozone O3, one O3 initial concentration inversion result is output by the ridge regression unit, one O3 initial concentration inversion result is output by the support vector machine unit, and one O3 initial concentration inversion result is output by the XGBoost unit, with a total of three O3 initial concentration inversion results.
And 2, determining a target concentration inversion result corresponding to the pixel point based on the first remote sensing image data, the first ground observation data, the first time-space data and each initial concentration inversion result through the second-stage combined sub-model. The second stage combination sub-model comprises two parts of breadth and depth, so that the space-time characteristics are extracted. In one embodiment, see steps 2.1 to 2.3 below:
And 2.1, performing independent heat coding treatment on the ecological geography type data and the landform type data through a linear regression network to obtain a sparse matrix, and determining a first concentration inversion result corresponding to the pixel points according to the first remote sensing image data, the first ground observation data, the sparse matrix and each initial concentration inversion result. In one embodiment, the unique thermal coding layer is used for carrying out unique thermal coding on classified data such as ecological geographic type data and landform type data to obtain a sparse matrix, and the linear regression layer is used for determining a first concentration inversion result corresponding to the pixel point according to continuous data such as first remote sensing image data and first ground observation data and combining the sparse matrix with each initial concentration inversion result.
In particular implementations, the breadth part is a linear regression model in which class variables (ecological geographic and geomorphic types) of spatiotemporal data are processed using one-hot encoding in the linear regression layer of the "breadth" part. The space-time data after the single thermal coding is changed into sparse matrix representation, so that the problem that the linear regression layer cannot benefit attribute data is solved, and the effect of expanding features is achieved to a certain extent.
And 2.2, performing dimension reduction processing on the ecological geography type data and the landform type data through a feedforward neural network to obtain a dense matrix, and determining a second concentration inversion result corresponding to the pixel points according to the first remote sensing image data, the first ground observation data, the dense matrix and each initial concentration inversion result. In one embodiment, the embedding layer is used for carrying out dimension reduction processing on classified data such as ecological geographic type data and landform type data to obtain a dense matrix, and the hidden layer is used for determining a second concentration inversion result corresponding to the pixel point according to continuous data such as first remote sensing image data and first ground observation data and combining the dense matrix with each initial concentration inversion result.
In a specific implementation, the depth portion employs an FFN model, and the "depth" portion is constructed as a dense matrix, i.e., an Embedding Layer (Embedding Layer), for the input class variables. The embedded layer can serve to reduce dimensions, similar toAnd the convolution operation reduces the occupation of subsequent resources. Each hidden layer contains one fully connected layer (FC), an exponential linear unit activation function (eLU), a Bulk Normalization (BN) and a dropout layer. eLU activation functions can avoid gradient extinction/explosion, converge faster, and have better generalization performance. In eLU activation functions, negative values will not be 0, effectively avoiding neuronal "death". The use of eLU can reduce training time and improve accuracy in neural networks compared to ReLU and variants thereof.
And 2.3, weighting the first concentration inversion result and the second concentration inversion result through a fusion network to obtain a target concentration inversion result corresponding to the pixel point. In one embodiment, the weight value used in the weighting process is also determined in the model training stage, and a weighted sum of the first concentration inversion result and the second concentration inversion result is calculated according to the weight value, where the weighted sum is the target concentration inversion result.
And step 3, obtaining a target concentration inversion result corresponding to the region to be researched based on the target concentration inversion result corresponding to each pixel point. In practical application, the steps 1 to 2 are repeatedly executed for each pixel point until inversion of all pixel points is completed, and a target concentration inversion result of ozone is output in a grid file format.
Further, the embodiment of the invention also provides a training step of the concentration inversion model cluster, wherein the training step can be seen in the following (1) to (5):
(1) Acquiring a second data set; the second data set comprises second remote sensing image data, second ground observation data, second space-time data and ground observation station pollutant data. Wherein the second data set, i.e., the data set used in the model training phase, ground observation station contaminant data is used as tag data (i.e., true values).
(2) The second data set is divided into a plurality of sub-data sets based on the land cover type in the second spatio-temporal data. For example, assuming that the pollutant is ozone, to overcome the influence of the inversion of the underlying surface type ozone, an initial concentration inversion model is trained according to the underlying surface land coverage type, and the land coverage type is classified according to water, building, desert, forest land, grassland, farmland and others according to the difference of ultraviolet reflection by the cover.
In a specific implementation, the second data set is overlaid with the land use type, and split into different sub-data sets separated according to the land cover type.
(3) For each sub-dataset, determining a predicted concentration inversion result based on the second remote sensing image data, the second ground observation data and the second spatiotemporal data in the sub-dataset by an initial concentration inversion model. In one embodiment, an initial predicted concentration inversion result may be determined by a first stage combined sub-model in the initial concentration inversion model based on the second remote sensing image data, the second ground observation data, and the second spatio-temporal data, and then a predicted concentration inversion result of the target may be determined by a second stage combined sub-model based on the second remote sensing image data, the second ground observation data, the second spatio-temporal data, and the initial predicted concentration inversion result. The inversion process can refer to the foregoing steps 1 to 3, which are not described in detail in the embodiment of the present invention.
(4) Training an initial concentration inversion model based on the pollutant data of the ground observation station in the sub-data set and the predicted concentration inversion result to obtain a target concentration inversion model corresponding to the sub-data set. In one embodiment, 60% of each sub-dataset may be used as a training dataset, 20% as a verification dataset, and 20% as a test dataset, the sub-datasets may be input into an initial concentration inversion model for training to obtain a target concentration inversion model for each sub-dataset, forming a concentration inversion model cluster for ozone inversion calculation.
(5) And combining the target concentration inversion model corresponding to each sub-data set into a concentration inversion model cluster.
In one embodiment, after determining the inversion result of the target concentration corresponding to the area to be studied, ozone high-value grid calculation may be further performed to realize visualization of the inversion result, specifically, target grid data corresponding to the area to be studied may be superimposed with the inversion result of the target concentration to determine a concentration value of the pollutant corresponding to each grid, or a concentration average value of the pollutant corresponding to each grid may be determined, and then the concentration value of the pollutant or the concentration average value of the pollutant may be sent to a designated association terminal, so that the concentration value of the pollutant or the concentration average value of the pollutant may be visually displayed through a graphical user interface of the designated association terminal.
In practical application, in order to meet the requirements of different users on the identification and display of the ozone high-value areas, the resampling method is used for obtaining the ozone high-value grid distribution with different resolutions. The method can be used for superposing the grid data of the target grids, calculating the ozone value in the grids grid by grid, and is suitable for the situation that the resolution of the target grids is higher or lower than the ozone inversion spatial resolution, and calculating non-square grids or irregular grids according to the needs.
The step of calculating the ith grid ozone value is as follows:
(1) Obtaining an ozone inversion result pixel set contained in or intersected by a single grid{/>,j=1,2,...n}，Wherein->For picture elements, n is the number of picture elements, +.>Is the pixel area->Is the mean value of the pixel ozone.
(2) Calculating grid ozone mean value：
The method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>Is the mesh area.
In summary, the embodiment of the invention inverts the atmospheric ozone concentration by combining the high-resolution five-star atmospheric trace gas differential absorption spectrometer (EMI) hyperspectral data with ground observation data based on different land coverage types, thereby overcoming inversion errors caused by the assumption of the traditional model on an ideal underlying surface, improving inversion precision and inversion efficiency. The ozone inversion result is subjected to gridding treatment, and different resolution ratios and different shape grids can be divided according to different management layer requirements, so that the ozone pollution management and tracing requirements of different administrative areas are met, visual display can be more intuitively provided for management staff, and a basis is provided for finding key pollution areas.
For easy understanding, the embodiment of the invention takes inversion of ozone concentration as an example, provides an application example of a pollutant concentration inversion method, selects 2021, 11, 6 and 6 days GF5B/EMI (UV 2) data as original observation data, and the specific implementation area is a large area of Yunnan province in an observation range, and the specific range is shown in an observation coverage schematic diagram shown in fig. 3. Wherein, (3 a) is the observation coverage area of Yunnan province, and (3 b) is the observation station in the coverage area.
Further, referring to a schematic diagram of ozone observation effect shown in fig. 4, wherein (4 a) is an observed value of O3 concentration of a ground station, (4B) is an estimated value of concentration of a corresponding station, (4 c) is an estimated value of GF5B/EMI in a coverage area of yunnan province, (4 d) is a station-by-station verification result of the observed value and the estimated value, the overall correlation coefficient R2 is 0.52, and rmse is 8.77 μg/m3.
In summary, the method for inverting the concentration of the pollutant provided by the embodiment of the invention can well invert the concentration of ozone, and the embodiment of the invention has at least the following characteristics:
(1) The near-surface O3 concentration is directly inverted through the hyperspectral observation satellite, and accumulated errors caused by indirect inversion are reduced.
(2) A near-ground O3 concentration inversion algorithm aiming at a hyperspectral observation satellite is constructed, and the algorithm combines a deep learning model and a linear regression model, so that the inversion accuracy is improved, and meanwhile, the interpretability is improved.
(3) According to the land coverage types of the underlying surfaces, inversion model clusters are trained respectively, so that the models are more targeted, the influence of different underlying surfaces on ultraviolet reflection can be overcome, the influence of the ideal underlying surface assumption of the existing model on inversion precision is overcome, and ozone precision can be improved.
(4) The ozone inversion result is subjected to gridding treatment, different resolution ratios and different shape grids are divided according to different management layer requirements, the ozone pollution management and tracing requirements of different administrative areas are met, visual display can be more intuitively provided for management staff, and a basis is provided for finding out important pollution areas.
For the method for inverting the concentration of the contaminant provided in the foregoing embodiment, the embodiment of the present invention provides an apparatus for inverting the concentration of the contaminant, referring to a schematic structural diagram of the apparatus for inverting the concentration of the contaminant shown in fig. 5, the apparatus mainly includes the following parts:
the data acquisition module 502 is configured to acquire a first data set corresponding to a region to be studied; the first data set comprises first remote sensing image data, first ground observation data and first time-space data;
the model determining module 504 is configured to determine, according to the land coverage type in the first time-space data, a target concentration inversion model corresponding to each pixel point in each region to be studied from concentration inversion model clusters obtained by training in advance; the concentration inversion model cluster comprises concentration inversion models corresponding to various land coverage types;
And the concentration inversion module 506 is configured to invert the concentration of the pollutant in the region to be studied based on the first data set through each target concentration inversion model, so as to obtain a target concentration inversion result corresponding to the region to be studied.
According to the pollutant concentration inversion device provided by the embodiment of the invention, aiming at the first remote sensing image data, the first ground observation data and the first time space data are combined, meanwhile, the pollutant concentration in the area to be researched is inverted by utilizing the corresponding target concentration inversion model in the concentration inversion model cluster based on the land coverage types in different first time space data, and the inversion error caused by the assumption of the ideal underlying surface by the traditional model is overcome, so that the complexity of the pollutant concentration inversion can be remarkably reduced, the inversion efficiency of the pollutant concentration can be effectively improved, and the inversion precision of the pollutant concentration can be remarkably improved.
In one embodiment, the target concentration inversion model includes a first stage combining sub-model and a second stage combining sub-model; the concentration inversion module 506 is also configured to:
determining at least one initial concentration inversion result corresponding to the pixel point based on the first remote sensing image data, the first ground observation data and the first time-space data through the first-stage combined sub-model; the first-stage combined sub-model comprises at least one inversion unit, wherein each inversion unit is used for determining an initial concentration inversion result corresponding to a pixel point;
Determining a target concentration inversion result corresponding to the pixel point based on the first remote sensing image data, the first ground observation data, the first time-space data and each initial concentration inversion result through a second stage combined sub-model;
and obtaining a target concentration inversion result corresponding to the region to be researched based on the target concentration inversion result corresponding to each pixel point.
In one embodiment, the second stage combining sub-model includes a linear regression network, a feed forward neural network, and a fusion network, and the first time space data further includes ecological geography type data and relief type data; the concentration inversion module 506 is also configured to:
performing independent heat coding on the ecological geographic type data and the geomorphic type data through a linear regression network to obtain a sparse matrix, and determining a first concentration inversion result corresponding to the pixel points according to the first remote sensing image data, the first ground observation data, the sparse matrix and each initial concentration inversion result;
performing dimension reduction processing on the ecological geography type data and the landform type data through a feedforward neural network to obtain a dense matrix, and determining a second concentration inversion result corresponding to the pixel points according to the first remote sensing image data, the first ground observation data, the dense matrix and each initial concentration inversion result;
And weighting the first concentration inversion result and the second concentration inversion result through a fusion network to obtain a target concentration inversion result corresponding to the pixel point.
In one embodiment, the inversion unit is one or more of an XGBoost unit, a ridge regression unit, and a support vector machine unit.
In one embodiment, the training module is further configured to:
acquiring a second data set; the second data set comprises second remote sensing image data, second ground observation data, second space-time data and ground observation station pollutant data;
dividing the second data set into a plurality of sub-data sets based on the land cover type in the second spatio-temporal data;
for each sub-data set, determining a predicted concentration inversion result based on second remote sensing image data, second ground observation data and second space-time data in the sub-data set by an initial concentration inversion model;
training an initial concentration inversion model based on pollutant data and predicted concentration inversion results of a ground observation station in the sub-data set to obtain a target concentration inversion model corresponding to the sub-data set;
and combining the target concentration inversion model corresponding to each sub-data set into a concentration inversion model cluster.
In one embodiment, the concentration inversion module 506 is further configured to:
performing radiation calibration on the first remote sensing image data to obtain a calibrated radiation brightness matrix;
zero-equalizing the calibrated radiation brightness matrix to obtain an equalized radiation brightness matrix;
determining a covariance matrix based on the calibrated radiation brightness matrix and the averaged radiation brightness matrix, and performing data dimension reduction processing based on the eigenvalues and eigenvectors of the covariance matrix to obtain dimension reduced data;
and carrying out pollutant concentration inversion on the region to be researched based on the dimensionality reduced data, the first ground observation data and the first time space data through each target concentration inversion model to obtain a target concentration inversion result corresponding to the region to be researched.
In one embodiment, the method further comprises a visualization module for:
superposing target grid data corresponding to the region to be researched and a target concentration inversion result to determine a pollutant concentration value corresponding to each grid or determine a pollutant concentration average value corresponding to each grid;
and sending the pollutant concentration value or the pollutant concentration mean value to the appointed association terminal so as to visually display the pollutant concentration value or the pollutant concentration mean value through a graphical user interface of the appointed association terminal.
The device provided by the embodiment of the present invention has the same implementation principle and technical effects as those of the foregoing method embodiment, and for the sake of brevity, reference may be made to the corresponding content in the foregoing method embodiment where the device embodiment is not mentioned.
The embodiment of the invention provides electronic equipment, which comprises a processor and a storage device; the storage means has stored thereon a computer program which, when executed by the processor, performs the method of any of the embodiments described above.
Fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present invention, where the electronic device 100 includes: a processor 60, a memory 61, a bus 62 and a communication interface 63, the processor 60, the communication interface 63 and the memory 61 being connected by the bus 62; the processor 60 is arranged to execute executable modules, such as computer programs, stored in the memory 61.
The memory 61 may include a high-speed random access memory (RAM, random Access Memory), and may further include a non-volatile memory (non-volatile memory), such as at least one magnetic disk memory. The communication connection between the system network element and at least one other network element is achieved via at least one communication interface 63 (which may be wired or wireless), and may use the internet, a wide area network, a local network, a metropolitan area network, etc.
Bus 62 may be an ISA bus, a PCI bus, an EISA bus, or the like. The buses may be classified as address buses, data buses, control buses, etc. For ease of illustration, only one bi-directional arrow is shown in FIG. 6, but not only one bus or type of bus.
The memory 61 is configured to store a program, and the processor 60 executes the program after receiving an execution instruction, and the method executed by the apparatus for flow defining disclosed in any of the foregoing embodiments of the present invention may be applied to the processor 60 or implemented by the processor 60.
The processor 60 may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method may be performed by integrated logic circuitry in hardware or instructions in software in the processor 60. The processor 60 may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), and the like; but may also be a digital signal processor (Digital Signal Processing, DSP for short), application specific integrated circuit (Application Specific Integrated Circuit, ASIC for short), off-the-shelf programmable gate array (Field-Programmable Gate Array, FPGA for short), or other programmable logic device, discrete gate or transistor logic device, discrete hardware components. The disclosed methods, steps, and logic blocks in the embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in connection with the embodiments of the present invention may be embodied directly in the execution of a hardware decoding processor, or in the execution of a combination of hardware and software modules in a decoding processor. The software modules may be located in a random access memory, flash memory, read only memory, programmable read only memory, or electrically erasable programmable memory, registers, etc. as well known in the art. The storage medium is located in a memory 61 and the processor 60 reads the information in the memory 61 and in combination with its hardware performs the steps of the method described above.
The computer program product of the readable storage medium provided by the embodiment of the present invention includes a computer readable storage medium storing a program code, where the program code includes instructions for executing the method described in the foregoing method embodiment, and the specific implementation may refer to the foregoing method embodiment and will not be described herein.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
Finally, it should be noted that: the above examples are only specific embodiments of the present invention, and are not intended to limit the scope of the present invention, but it should be understood by those skilled in the art that the present invention is not limited thereto, and that the present invention is described in detail with reference to the foregoing examples: any person skilled in the art may modify or easily conceive of the technical solution described in the foregoing embodiments, or perform equivalent substitution of some of the technical features, while remaining within the technical scope of the present disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention, and are intended to be included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A method of contaminant concentration inversion comprising:
acquiring a first data set corresponding to a region to be researched; the first data set comprises first remote sensing image data, first ground observation data and first time-space data;
According to the land coverage type in the first time-space data, determining a target concentration inversion model corresponding to each pixel point in each region to be researched from a concentration inversion model cluster obtained through pre-training; the concentration inversion model cluster comprises concentration inversion models corresponding to various land coverage types;
and carrying out pollutant concentration inversion on the region to be researched based on the first data set through each target concentration inversion model to obtain a target concentration inversion result corresponding to the region to be researched.
2. The contaminant concentration inversion method of claim 1, wherein the target concentration inversion model comprises a first stage combination sub-model and a second stage combination sub-model;
carrying out pollutant concentration inversion on the region to be researched based on the first data set through each target concentration inversion model to obtain a target concentration inversion result corresponding to the region to be researched, wherein the method comprises the following steps:
determining at least one initial concentration inversion result corresponding to the pixel point based on the first remote sensing image data, the first ground observation data and the first time-space data through the first stage combination sub-model; the first-stage combined sub-model comprises at least one inversion unit, wherein each inversion unit is used for determining an initial concentration inversion result corresponding to the pixel point;
Determining a target concentration inversion result corresponding to the pixel point based on the first remote sensing image data, the first ground observation data, the first time-space data and each initial concentration inversion result through the second-stage combined sub-model;
and obtaining the target concentration inversion result corresponding to the region to be researched based on the target concentration inversion result corresponding to each pixel point.
3. The contaminant concentration inversion method of claim 2, wherein the second stage combining sub-model comprises a linear regression network, a feed forward neural network, and a fusion network, the first time-space data further comprising ecology geographic type data and relief type data;
determining, by the second stage combined sub-model, a target concentration inversion result corresponding to the pixel point based on the first remote sensing image data, the first ground observation data, the first time-space data, and each of the initial concentration inversion results, including:
performing single-heat encoding processing on the ecological geography type data and the landform type data through the linear regression network to obtain a sparse matrix, and determining a first concentration inversion result corresponding to the pixel point according to the first remote sensing image data, the first ground observation data, the sparse matrix and each initial concentration inversion result;
Performing dimension reduction processing on the ecological geography type data and the landform type data through the feedforward neural network to obtain a dense matrix, and determining a second concentration inversion result corresponding to the pixel point according to the first remote sensing image data, the first ground observation data, the dense matrix and each initial concentration inversion result;
and weighting the first concentration inversion result and the second concentration inversion result through the fusion network to obtain a target concentration inversion result corresponding to the pixel point.
4. The contaminant concentration inversion method of claim 2, wherein the inversion unit is one or more of an XGBoost unit, a ridge regression unit, and a support vector machine unit.
5. The method of claim 1, wherein the training step of the concentration inversion model cluster comprises:
acquiring a second data set; the second data set comprises second remote sensing image data, second ground observation data, second space-time data and ground observation station pollutant data;
dividing the second data set into a plurality of sub-data sets based on a land cover type in the second spatio-temporal data;
For each of the sub-data sets, determining a predicted concentration inversion result based on the second remote sensing image data, the second ground observation data, and the second spatiotemporal data in the sub-data set by an initial concentration inversion model;
training the initial concentration inversion model based on the pollutant data of the ground observation station in the sub-data set and the predicted concentration inversion result to obtain a target concentration inversion model corresponding to the sub-data set;
and combining the target concentration inversion models corresponding to each sub-data set into a concentration inversion model cluster.
6. The contaminant concentration inversion method according to claim 1, wherein the target concentration inversion result corresponding to the region to be studied is obtained by inverting the contaminant concentration of the region to be studied based on the first data set, further comprising:
performing radiation calibration on the first remote sensing image data to obtain a calibrated radiation brightness matrix;
zero-equalizing the scaled radiation brightness matrix to obtain an equalized radiation brightness matrix;
determining a covariance matrix based on the scaled radiation brightness matrix and the averaged radiation brightness matrix, and performing data dimension reduction processing based on the eigenvalues and eigenvectors of the covariance matrix to obtain dimension reduced data;
And carrying out pollutant concentration inversion on the region to be researched based on the dimension-reduced data, the first ground observation data and the first time space data through each target concentration inversion model to obtain a target concentration inversion result corresponding to the region to be researched.
7. The contaminant concentration inversion method of claim 1, wherein after inverting the contaminant concentration of the region to be studied based on the first data set to obtain a target concentration inversion result corresponding to the region to be studied, the method further comprises:
superposing target grid data corresponding to the region to be researched and the target concentration inversion result to determine a pollutant concentration value corresponding to each grid or determine a pollutant concentration average value corresponding to each grid;
and sending the pollutant concentration value or the pollutant concentration mean value to a designated associated terminal so as to visually display the pollutant concentration value or the pollutant concentration mean value through a graphic user interface of the designated associated terminal.
8. A contaminant concentration inversion apparatus, comprising:
the data acquisition module is used for acquiring a first data set corresponding to the region to be researched; the first data set comprises first remote sensing image data, first ground observation data and first time-space data;
The model determining module is used for determining a target concentration inversion model corresponding to each pixel point in each region to be researched from concentration inversion model clusters obtained through pre-training according to the land coverage type in the first time-space data; the concentration inversion model cluster comprises concentration inversion models corresponding to various land coverage types;
and the concentration inversion module is used for inverting the concentration of the pollutant in the region to be researched based on the first data set through each target concentration inversion model to obtain a target concentration inversion result corresponding to the region to be researched.
9. An electronic device comprising a processor and a memory, the memory storing computer-executable instructions executable by the processor, the processor executing the computer-executable instructions to implement the method of any one of claims 1 to 7.
10. A computer readable storage medium storing computer executable instructions which, when invoked and executed by a processor, cause the processor to implement the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310770616.6A CN116504330B (en) | 2023-06-28 | 2023-06-28 | Pollutant concentration inversion method and device, electronic equipment and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310770616.6A CN116504330B (en) | 2023-06-28 | 2023-06-28 | Pollutant concentration inversion method and device, electronic equipment and readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116504330A true CN116504330A (en) | 2023-07-28 |
| CN116504330B CN116504330B (en) | 2023-09-19 |
Family
ID=87321668
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310770616.6A Active CN116504330B (en) | 2023-06-28 | 2023-06-28 | Pollutant concentration inversion method and device, electronic equipment and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116504330B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116910457A (en) * | 2023-09-13 | 2023-10-20 | 北京师范大学 | Pollutant inversion method and device based on region |
| CN116994146A (en) * | 2023-09-26 | 2023-11-03 | 航天宏图信息技术股份有限公司 | Near-ground pollutant short-term prediction method, device and equipment based on SL4DID |
| CN117147778A (en) * | 2023-10-31 | 2023-12-01 | 航天宏图信息技术股份有限公司 | Method and device for tracing and monitoring atmospheric pollutants, electronic equipment and storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109001091A (en) * | 2018-07-18 | 2018-12-14 | 北京航天宏图信息技术股份有限公司 | Satellite remote-sensing monitoring method, device and the computer-readable medium of atmosphere pollution |
| CN109784552A (en) * | 2018-12-29 | 2019-05-21 | 武汉大学 | A kind of construction method of the space variable coefficient PM2.5 concentration appraising model based on Re-ESF algorithm |
| WO2021208393A1 (en) * | 2020-04-15 | 2021-10-21 | 北京工业大学 | Inversion estimation method for air pollutant emission inventory |
| CN114357894A (en) * | 2022-03-11 | 2022-04-15 | 阿里巴巴达摩院(杭州)科技有限公司 | Atmospheric pollutant processing method, storage medium and computer terminal |
| CN114926749A (en) * | 2022-07-22 | 2022-08-19 | 山东大学 | Near-surface atmospheric pollutant inversion method and system based on remote sensing image |
| CN115420690A (en) * | 2022-04-29 | 2022-12-02 | 中遥环境(西安)股份有限公司 | Near-surface trace gas concentration inversion model and inversion method |
| CN116340863A (en) * | 2023-05-30 | 2023-06-27 | 航天宏图信息技术股份有限公司 | Air pollutant prediction method and device, electronic equipment and readable storage medium |
-
2023
- 2023-06-28 CN CN202310770616.6A patent/CN116504330B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109001091A (en) * | 2018-07-18 | 2018-12-14 | 北京航天宏图信息技术股份有限公司 | Satellite remote-sensing monitoring method, device and the computer-readable medium of atmosphere pollution |
| CN109784552A (en) * | 2018-12-29 | 2019-05-21 | 武汉大学 | A kind of construction method of the space variable coefficient PM2.5 concentration appraising model based on Re-ESF algorithm |
| WO2021208393A1 (en) * | 2020-04-15 | 2021-10-21 | 北京工业大学 | Inversion estimation method for air pollutant emission inventory |
| CN114357894A (en) * | 2022-03-11 | 2022-04-15 | 阿里巴巴达摩院(杭州)科技有限公司 | Atmospheric pollutant processing method, storage medium and computer terminal |
| CN115420690A (en) * | 2022-04-29 | 2022-12-02 | 中遥环境(西安)股份有限公司 | Near-surface trace gas concentration inversion model and inversion method |
| CN114926749A (en) * | 2022-07-22 | 2022-08-19 | 山东大学 | Near-surface atmospheric pollutant inversion method and system based on remote sensing image |
| CN116340863A (en) * | 2023-05-30 | 2023-06-27 | 航天宏图信息技术股份有限公司 | Air pollutant prediction method and device, electronic equipment and readable storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 蔡建楠 等: "高分一、二号卫星遥感数据在生态环境监测中的应用", 环境监控与预警, pages 16 - 22 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116910457A (en) * | 2023-09-13 | 2023-10-20 | 北京师范大学 | Pollutant inversion method and device based on region |
| CN116910457B (en) * | 2023-09-13 | 2023-12-08 | 北京师范大学 | Pollutant inversion method and device based on region |
| CN116994146A (en) * | 2023-09-26 | 2023-11-03 | 航天宏图信息技术股份有限公司 | Near-ground pollutant short-term prediction method, device and equipment based on SL4DID |
| CN116994146B (en) * | 2023-09-26 | 2023-12-29 | 航天宏图信息技术股份有限公司 | Near-ground pollutant short-term prediction method, device and equipment based on SL4DID |
| CN117147778A (en) * | 2023-10-31 | 2023-12-01 | 航天宏图信息技术股份有限公司 | Method and device for tracing and monitoring atmospheric pollutants, electronic equipment and storage medium |
| CN117147778B (en) * | 2023-10-31 | 2024-01-05 | 航天宏图信息技术股份有限公司 | Method and device for tracing and monitoring atmospheric pollutants, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116504330B (en) | 2023-09-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116504330B (en) | Pollutant concentration inversion method and device, electronic equipment and readable storage medium | |
| Kang et al. | Estimation of surface-level NO2 and O3 concentrations using TROPOMI data and machine learning over East Asia | |
| Hemer et al. | Projected changes in wave climate from a multi-model ensemble | |
| Li et al. | Predicting ground-level PM2. 5 concentrations in the Beijing-Tianjin-Hebei region: A hybrid remote sensing and machine learning approach | |
| CN113128134B (en) | Mining area ecological environment evolution driving factor weight quantitative analysis method | |
| CN111007021A (en) | Hyperspectral water quality parameter inversion system and method based on one-dimensional convolution neural network | |
| Alamgir et al. | Downscaling and projection of spatiotemporal changes in temperature of Bangladesh | |
| Cho et al. | A novel ensemble learning for post-processing of NWP Model's next-day maximum air temperature forecast in summer using deep learning and statistical approaches | |
| Ares et al. | Detection of process‐related changes in plant patterns at extended spatial scales during early dryland desertification | |
| CN112052627A (en) | Method, device, medium and equipment for estimating near-surface ozone space distribution | |
| Wahla et al. | Mapping and monitoring of spatio-temporal land use and land cover changes and relationship with normalized satellite indices and driving factors | |
| Yeom et al. | Monitoring paddy productivity in North Korea employing geostationary satellite images integrated with GRAMI-rice model | |
| Peng et al. | Global estimates of 500 m daily aerodynamic roughness length from MODIS data | |
| Li et al. | Estimation of aboveground vegetation biomass based on Landsat-8 OLI satellite images in the Guanzhong Basin, China | |
| Scheibenreif et al. | Toward global estimation of ground-level no 2 pollution with deep learning and remote sensing | |
| Murthy et al. | Paddy crop insurance using satellite-based composite index of crop performance | |
| Briegel et al. | Modelling long-term thermal comfort conditions in urban environments using a deep convolutional encoder-decoder as a computational shortcut | |
| Rumapea et al. | Improving Convective Cloud Classification with Deep Learning: The CC-Unet Model. | |
| CN116340863B (en) | Air pollutant prediction method and device, electronic equipment and readable storage medium | |
| Kuter et al. | Modern applied mathematics for alternative modeling of the atmospheric effects on satellite images | |
| Sun et al. | Leaf area index remote sensing based on Deep Belief Network supported by simulation data | |
| Lee et al. | Surface albedo from the geostationary Communication, Ocean and Meteorological Satellite (COMS)/Meteorological Imager (MI) observation system | |
| Chen et al. | Remote sensing retrieval of aerosol types in China using geostationary satellite | |
| Tang et al. | A review of machine learning for modeling air quality: Overlooked but important issues | |
| Wang et al. | Generalized fine-resolution FPAR estimation using Google Earth Engine: Random forest or multiple linear regression |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |